學術產出-Theses
Article View/Open
Publication Export
-
題名 基於孿生網絡之正則化對比式遷移學習於醫療影像
Contrastive Transfer Learning for Regularization with Triplet Network on Medical Imaging作者 游勤葑
Yu, Chin-Feng貢獻者 邱淑怡
Chiu, Shu-I
游勤葑
Yu, Chin-Feng關鍵詞 黃斑部病變
對比式學習
遷移式學習
正則化
Macular degeneration
Contrastive learning
Transfer learning
Regularization日期 2022 上傳時間 5-Oct-2022 09:15:57 (UTC+8) 摘要 在此篇論文中,我們針對眼底攝影 ( Color Fundus Photography)醫療影像提出了一個新穎的遷移式學習架構,稱為基於孿生網絡之正則化對比式遷移學習(Contrastive Transfer Learning for Regularization with Triplet Network),CTLRT,在 CTLRT 中包含三種對比式正則化損失項且結合了遷移式學習的骨架,我們在三種眼底攝影資料集且多種遷移式學習骨架下表明 CTLRT 不只擁有比傳統的遷移式學習更高的準確度,並且透過我們設計的對比式正則化損失減緩複雜模型帶來的過擬合效應,提高了模型的泛化能力,且經由可視化模型關注的區域說明了 CTLRT 確實能正確的關注變病的區域。
This paper focuses on Color Fundus Photography and proposes a novel transfer learning architecture called Contrastive Transfer Learning for Regularization with Triplet Network (CTLRT). CTLRT contains three kinds of contrastive regularization loss terms and combines the backbone of transfer learning. We use three fundus photography datasets and multiple transfer backbones. The following shows that CTLRT not only has higher accuracy than traditional transfer learning but also mitigates the overfitting effect brought by complex models through our designed contrastive regularizationloss and improves the model’s generalization ability. Visualizing the area where model interest shows that CTLRT correctly focuses on the diseased site.參考文獻 Agarap, A. F. (2018). Deep learning using rectified linear units (relu). arXiv preprintarXiv:1803.08375.Bromley, J., Guyon, I., LeCun, Y., Säckinger, E., and Shah, R. (1993). Signature verifi-cation using a” siamese” time delay neural network. Advances in neural informationprocessing systems, 6.Chakraborty, R. and Pramanik, A. (2022). Dcnn-based prediction model for detectionof age-related macular degeneration from color fundus images. Medical & BiologicalEngineering & Computing, 60(5):1431–1448.Chen, T., Kornblith, S., Norouzi, M., and Hinton, G. (2020a). A simple framework forcontrastive learning of visual representations. In International conference on machinelearning, pages 1597–1607. PMLR.Chen, T.-C., Lim, W. S., Wang, V. Y., Ko, M.-L., Chiu, S.-I., Huang, Y.-S., Lai, F., Yang,C.-M., Hu, F.-R., Jang, J.-S. R., et al. (2021). Artificial intelligence–assisted early detec-tion of retinitis pigmentosa—the most common inherited retinal degeneration. Journalof Digital Imaging, 34(4):948–958.Chen, X., Fan, H., Girshick, R., and He, K. (2020b). Improved baselines with momentumcontrastive learning. arXiv preprint arXiv:2003.04297.Chen, X. and He, K. (2021). Exploring simple siamese representation learning. In Pro-ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,pages 15750–15758.Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. InProceedings of the IEEE conference on computer vision and pattern recognition, pages1251–1258.Dataset, B. R. O.-A. (2019). ichallenge-amd.Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009a). Imagenet: Alarge-scale hierarchical image database. In 2009 IEEE Conference on Computer Visionand Pattern Recognition, pages 248–255.Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009b). ImageNet: ALarge-Scale Hierarchical Image Database. In CVPR09.Farnell, D. J., Hatfield, F. N., Knox, P., Reakes, M., Spencer, S., Parry, D., and Harding, S. P. (2008). Enhancement of blood vessels in digital fundus photographs via the appli-cation of multiscale line operators. Journal of the Franklin institute, 345(7):748–765.Fukushima, K. and Miyake, S. (1982). Neocognitron: A new algorithm for pattern recog-nition tolerant of deformations and shifts in position. Pattern recognition, 15(6):455–469.Grill, J.-B., Strub, F., Altché, F., Tallec, C., Richemond, P., Buchatskaya, E., Doersch, C.,Avila Pires, B., Guo, Z., Gheshlaghi Azar, M., et al. (2020). Bootstrap your own latent-anew approach to self-supervised learning. Advances in Neural Information ProcessingSystems, 33:21271–21284.Hadsell, R., Chopra, S., and LeCun, Y. (2006). Dimensionality reduction by learning aninvariant mapping. In 2006 IEEE Computer Society Conference on Computer Visionand Pattern Recognition (CVPR’06), volume 2, pages 1735–1742. IEEE.He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recogni-tion. In Proceedings of the IEEE conference on computer vision and pattern recogni-tion, pages 770–778.Huang, G., Liu, Z., Van Der Maaten, L., and Weinberger, K. Q. (2017). Densely connectedconvolutional networks. In Proceedings of the IEEE conference on computer vision andpattern recognition, pages 4700–4708.Hubel, D. H. and Wiesel, T. N. (1962). Receptive fields, binocular interaction and func-tional architecture in the cat’s visual cortex. The Journal of physiology, 160(1):106.Krizhevsky, A., Hinton, G., et al. (2009). Learning multiple layers of features from tinyimages. Technical report, Citeseer.Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012). Imagenet classification with deepconvolutional neural networks. In Advances in neural information processing systems,pages 1097–1105.LeCun, Y., Boser, B., Denker, J., Henderson, D., Howard, R., Hubbard, W., and Jackel, L.(1989). Handwritten digit recognition with a back-propagation network. Advances in neural information processing systems, 2.LeCun, Y., Bottou, L., Bengio, Y., Haffner, P., et al. (1998). Gradient-based learningapplied to document recognition. Proceedings of the IEEE, 86(11):2278–2324.Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., and Ng, A. Y. (2011). Readingdigits in natural images with unsupervised feature learning.Pan, S. J. and Yang, Q. (2009). A survey on transfer learning. IEEE Transactions onknowledge and data engineering, 22(10):1345–1359.Raghu, M., Zhang, C., Kleinberg, J., and Bengio, S. (2019). Transfusion: Understand-ing transfer learning for medical imaging. Advances in neural information processingsystems, 32.Robinson, J., Chuang, C.-Y., Sra, S., and Jegelka, S. (2020). Contrastive learning withhard negative samples. arXiv preprint arXiv:2010.04592.Rumelhart, D. E., Hinton, G. E., and Williams, R. J. (1986). Learning representations byback-propagating errors. nature, 323(6088):533–536.Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., and Batra, D. (2017). Grad-cam: Visual explanations from deep networks via gradient-based localization. InProceedings of the IEEE international conference on computer vision, pages 618–626.Shorten, C. and Khoshgoftaar, T. M. (2019). A survey on image data augmentation fordeep learning. Journal of Big Data, 6(1):60.Smith, R. (2007). An overview of the tesseract ocr engine. In Ninth international confer-ence on document analysis and recognition (ICDAR 2007), volume 2, pages 629–633.IEEE.Szegedy, C., Ioffe, S., Vanhoucke, V., and Alemi, A. A. (2017). Inception-v4, inception- resnet and the impact of residual connections on learning. In AAAI, volume 4, page 12.Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., and Van-houcke (2015). Going deeper with convolutions. In Proceedings of the IEEE conferenceon computer vision and pattern recognition, pages 1–9.Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., and Wojna, Z. (2016). Rethinking theinception architecture for computer vision. In Proceedings of the IEEE conference oncomputer vision and pattern recognition, pages 2818–2826.Torrey, L. and Shavlik, J. (2010). Transfer learning. In Handbook of research on machinelearning applications and trends: algorithms, methods, and techniques, pages 242–264.IGI global.Yu, Y., Chen, X., Zhu, X., Zhang, P., Hou, Y., Zhang, R., and Wu, C. (2020). Performanceof deep transfer learning for detecting abnormal fundus images. Journal of CurrentOphthalmology, 32(4):368.Zhai, X., Oliver, A., Kolesnikov, A., and Beyer, L. (2019). S4l: Self-supervised semi-supervised learning. In Proceedings of the IEEE/CVF International Conference onComputer Vision, pages 1476–1485. 描述 碩士
國立政治大學
資訊科學系
110753205資料來源 http://thesis.lib.nccu.edu.tw/record/#G0110753205 資料類型 thesis dc.contributor.advisor 邱淑怡 zh_TW dc.contributor.advisor Chiu, Shu-I en_US dc.contributor.author (Authors) 游勤葑 zh_TW dc.contributor.author (Authors) Yu, Chin-Feng en_US dc.creator (作者) 游勤葑 zh_TW dc.creator (作者) Yu, Chin-Feng en_US dc.date (日期) 2022 en_US dc.date.accessioned 5-Oct-2022 09:15:57 (UTC+8) - dc.date.available 5-Oct-2022 09:15:57 (UTC+8) - dc.date.issued (上傳時間) 5-Oct-2022 09:15:57 (UTC+8) - dc.identifier (Other Identifiers) G0110753205 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/142128 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 資訊科學系 zh_TW dc.description (描述) 110753205 zh_TW dc.description.abstract (摘要) 在此篇論文中,我們針對眼底攝影 ( Color Fundus Photography)醫療影像提出了一個新穎的遷移式學習架構,稱為基於孿生網絡之正則化對比式遷移學習(Contrastive Transfer Learning for Regularization with Triplet Network),CTLRT,在 CTLRT 中包含三種對比式正則化損失項且結合了遷移式學習的骨架,我們在三種眼底攝影資料集且多種遷移式學習骨架下表明 CTLRT 不只擁有比傳統的遷移式學習更高的準確度,並且透過我們設計的對比式正則化損失減緩複雜模型帶來的過擬合效應,提高了模型的泛化能力,且經由可視化模型關注的區域說明了 CTLRT 確實能正確的關注變病的區域。 zh_TW dc.description.abstract (摘要) This paper focuses on Color Fundus Photography and proposes a novel transfer learning architecture called Contrastive Transfer Learning for Regularization with Triplet Network (CTLRT). CTLRT contains three kinds of contrastive regularization loss terms and combines the backbone of transfer learning. We use three fundus photography datasets and multiple transfer backbones. The following shows that CTLRT not only has higher accuracy than traditional transfer learning but also mitigates the overfitting effect brought by complex models through our designed contrastive regularizationloss and improves the model’s generalization ability. Visualizing the area where model interest shows that CTLRT correctly focuses on the diseased site. en_US dc.description.tableofcontents 摘要 iAbstract ii目錄 iii圖目錄 v表目錄 vii第 一章 緒論 11.1 研究背景與動機 11.2 研究問題與目的 31.3 論文架構 5第 二章 文獻探討 62.1 深度卷積神經網絡 62.2 深度卷積神經網路與醫療影像 72.3 遷移式學習與醫療影像 72.4 資料增強 82.5 對比式學習 102.6 自監督式學習 122.6.1 探索簡單的孿生表達學習 12第 三章 研究方法 143.1 基於孿生網絡之正則化對比式遷移學習 143.2 光學文字辨識 23第 四章 實驗分析 244.1 資料集 244.2 實驗設定及超參數設定 254.3 損失函數的訓練過程 264.4 CTLRT 以 Xception 為骨架 304.5 CTLRT 以 InceptionV3 為骨架 324.6 CTLRT 以 DenseNet201 為骨架 354.7 三種骨架之評估總結 384.8 可視化模型關注區域 394.9 ARIA and iChallenge-AMD 42第 五章 結論與未來展望 445.1 結論 445.2 未來展望 44參考文獻 46 zh_TW dc.format.extent 55359093 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0110753205 en_US dc.subject (關鍵詞) 黃斑部病變 zh_TW dc.subject (關鍵詞) 對比式學習 zh_TW dc.subject (關鍵詞) 遷移式學習 zh_TW dc.subject (關鍵詞) 正則化 zh_TW dc.subject (關鍵詞) Macular degeneration en_US dc.subject (關鍵詞) Contrastive learning en_US dc.subject (關鍵詞) Transfer learning en_US dc.subject (關鍵詞) Regularization en_US dc.title (題名) 基於孿生網絡之正則化對比式遷移學習於醫療影像 zh_TW dc.title (題名) Contrastive Transfer Learning for Regularization with Triplet Network on Medical Imaging en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) Agarap, A. F. (2018). Deep learning using rectified linear units (relu). arXiv preprintarXiv:1803.08375.Bromley, J., Guyon, I., LeCun, Y., Säckinger, E., and Shah, R. (1993). Signature verifi-cation using a” siamese” time delay neural network. Advances in neural informationprocessing systems, 6.Chakraborty, R. and Pramanik, A. (2022). Dcnn-based prediction model for detectionof age-related macular degeneration from color fundus images. Medical & BiologicalEngineering & Computing, 60(5):1431–1448.Chen, T., Kornblith, S., Norouzi, M., and Hinton, G. (2020a). A simple framework forcontrastive learning of visual representations. In International conference on machinelearning, pages 1597–1607. PMLR.Chen, T.-C., Lim, W. S., Wang, V. Y., Ko, M.-L., Chiu, S.-I., Huang, Y.-S., Lai, F., Yang,C.-M., Hu, F.-R., Jang, J.-S. R., et al. (2021). Artificial intelligence–assisted early detec-tion of retinitis pigmentosa—the most common inherited retinal degeneration. Journalof Digital Imaging, 34(4):948–958.Chen, X., Fan, H., Girshick, R., and He, K. (2020b). Improved baselines with momentumcontrastive learning. arXiv preprint arXiv:2003.04297.Chen, X. and He, K. (2021). Exploring simple siamese representation learning. In Pro-ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,pages 15750–15758.Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. InProceedings of the IEEE conference on computer vision and pattern recognition, pages1251–1258.Dataset, B. R. O.-A. (2019). ichallenge-amd.Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009a). Imagenet: Alarge-scale hierarchical image database. In 2009 IEEE Conference on Computer Visionand Pattern Recognition, pages 248–255.Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. (2009b). ImageNet: ALarge-Scale Hierarchical Image Database. In CVPR09.Farnell, D. J., Hatfield, F. N., Knox, P., Reakes, M., Spencer, S., Parry, D., and Harding, S. P. (2008). Enhancement of blood vessels in digital fundus photographs via the appli-cation of multiscale line operators. Journal of the Franklin institute, 345(7):748–765.Fukushima, K. and Miyake, S. (1982). Neocognitron: A new algorithm for pattern recog-nition tolerant of deformations and shifts in position. Pattern recognition, 15(6):455–469.Grill, J.-B., Strub, F., Altché, F., Tallec, C., Richemond, P., Buchatskaya, E., Doersch, C.,Avila Pires, B., Guo, Z., Gheshlaghi Azar, M., et al. (2020). Bootstrap your own latent-anew approach to self-supervised learning. Advances in Neural Information ProcessingSystems, 33:21271–21284.Hadsell, R., Chopra, S., and LeCun, Y. (2006). Dimensionality reduction by learning aninvariant mapping. In 2006 IEEE Computer Society Conference on Computer Visionand Pattern Recognition (CVPR’06), volume 2, pages 1735–1742. IEEE.He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image recogni-tion. In Proceedings of the IEEE conference on computer vision and pattern recogni-tion, pages 770–778.Huang, G., Liu, Z., Van Der Maaten, L., and Weinberger, K. Q. (2017). Densely connectedconvolutional networks. In Proceedings of the IEEE conference on computer vision andpattern recognition, pages 4700–4708.Hubel, D. H. and Wiesel, T. N. (1962). Receptive fields, binocular interaction and func-tional architecture in the cat’s visual cortex. The Journal of physiology, 160(1):106.Krizhevsky, A., Hinton, G., et al. (2009). Learning multiple layers of features from tinyimages. Technical report, Citeseer.Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012). Imagenet classification with deepconvolutional neural networks. In Advances in neural information processing systems,pages 1097–1105.LeCun, Y., Boser, B., Denker, J., Henderson, D., Howard, R., Hubbard, W., and Jackel, L.(1989). Handwritten digit recognition with a back-propagation network. Advances in neural information processing systems, 2.LeCun, Y., Bottou, L., Bengio, Y., Haffner, P., et al. (1998). Gradient-based learningapplied to document recognition. Proceedings of the IEEE, 86(11):2278–2324.Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., and Ng, A. Y. (2011). Readingdigits in natural images with unsupervised feature learning.Pan, S. J. and Yang, Q. (2009). A survey on transfer learning. IEEE Transactions onknowledge and data engineering, 22(10):1345–1359.Raghu, M., Zhang, C., Kleinberg, J., and Bengio, S. (2019). Transfusion: Understand-ing transfer learning for medical imaging. Advances in neural information processingsystems, 32.Robinson, J., Chuang, C.-Y., Sra, S., and Jegelka, S. (2020). Contrastive learning withhard negative samples. arXiv preprint arXiv:2010.04592.Rumelhart, D. E., Hinton, G. E., and Williams, R. J. (1986). Learning representations byback-propagating errors. nature, 323(6088):533–536.Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., and Batra, D. (2017). Grad-cam: Visual explanations from deep networks via gradient-based localization. InProceedings of the IEEE international conference on computer vision, pages 618–626.Shorten, C. and Khoshgoftaar, T. M. (2019). A survey on image data augmentation fordeep learning. Journal of Big Data, 6(1):60.Smith, R. (2007). An overview of the tesseract ocr engine. In Ninth international confer-ence on document analysis and recognition (ICDAR 2007), volume 2, pages 629–633.IEEE.Szegedy, C., Ioffe, S., Vanhoucke, V., and Alemi, A. A. (2017). Inception-v4, inception- resnet and the impact of residual connections on learning. In AAAI, volume 4, page 12.Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., and Van-houcke (2015). Going deeper with convolutions. In Proceedings of the IEEE conferenceon computer vision and pattern recognition, pages 1–9.Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., and Wojna, Z. (2016). Rethinking theinception architecture for computer vision. In Proceedings of the IEEE conference oncomputer vision and pattern recognition, pages 2818–2826.Torrey, L. and Shavlik, J. (2010). Transfer learning. In Handbook of research on machinelearning applications and trends: algorithms, methods, and techniques, pages 242–264.IGI global.Yu, Y., Chen, X., Zhu, X., Zhang, P., Hou, Y., Zhang, R., and Wu, C. (2020). Performanceof deep transfer learning for detecting abnormal fundus images. Journal of CurrentOphthalmology, 32(4):368.Zhai, X., Oliver, A., Kolesnikov, A., and Beyer, L. (2019). S4l: Self-supervised semi-supervised learning. In Proceedings of the IEEE/CVF International Conference onComputer Vision, pages 1476–1485. zh_TW dc.identifier.doi (DOI) 10.6814/NCCU202201567 en_US